理研計算科学研究センター長 Director RIKEN R-CCS, 東工大特定教授 Prof. Tokyo Tech. ACM/ISC/JSSST/IPSJ Fellows, IEEE Fernbach(2014)&Cray(2022) Awards, 令4紫綬褒章 Purple Ribbon Medal 2022
Following the latest AI boom, Fugaku's user support now uses LLM technology extensively:
富岳のユーザサポートも最先端のLLMテクノロジを活用しています。
English
R-CCS site:
【日本語】
理研サイト:
R-CCSサイト:
I am the author of this game, well much younger version of me during high school, written entirely in 6502 assembly, fitting within the 8KB RAM of the original PET. You might be able to guess my age…
I'd strongly support the idea of a Manhattan Project of intense research to make machines more trustworthy and interpretable (regardless of, or in parallel with a moratorium.) The premature super-investment in non-interpretable technologies is the core of our problems.
#Fugaku
has become No.1 in all the supercomputer performance benchmarks,
#Top500
,
#HPCG
, HPL-AI, and the
#Graph500
for the first time in history as a single machine simultaneously. Thanks for putting up the list!
There were unconfirmed info that HAL9000 “eye” was a Nikon Nikkor Fisheye Lens. I was happily able to confirm this. For 2001 geeks, It is a real movie prop with 18 rows of speaker perforations instead of ‘fake’ 12. So parts of HAL was made in Japan, as many good computers are :-)
The 'Myths' paper is out: Satoshi Matsuoka, Jens Domke, Mohamed Wahib, Aleksandr Drozd, Torsten Hoefler. "Myths and Legends in High-Performance Computing" " undergoing peer review. Hopefully the community will be engaged in yes/no discussions on Twitter!
Visiting ANL for a few days and got to see Aurora nearing completion. Don’t know if I can disclose the specs but it will become a supercomputer with largest # of GPUs ever, surpassing Frontier which has 40K. It will also likely be a machine with the highest power consumption.
We just announced the Fugaku config at a press briefing. The whole system is 158,976 A64FX nodes, and running at 2.2Ghz peak performances for 64bitFP/32bitFP/16bitFP/8bitINT are 537Peta/1.07Exa/2.15Exa/4.30Exa (FL)ops respectively, as well as the total memory BW 163 PetaByte/s.
We just had a joint press conference w/
#Suntory
&
#Toppan
regarding joint work in the development of a new face shield that will allow both
#COVID19
risk reduction & ease of eating/drinking, in restaurants, bars, parties, sports events,etc., based on
#Fugaku
droplet simulation.
Extremely honored to be awarded the 2022 IEEE-CS Seymour Cray Engineering Award, following the 2014 Fernbach Award. This is not an award for me, but rather for the achievements made by everyone involved throughout this incredible journey in
#HPC
.
Here is what
#Fugaku
looks like from up front. Yes, the obvious motif of Mt. Fuji, impressionist style, or "Kasumi-Fuji" (Mt. Fuji fogged by mist). Notice, out of the 12 rows with a front panel each, the center four do not have the Fugaku logo--- can someone guess why?
From April 1st I have become the Director of Riken Center for Computational Science, to lead the K-Computer & Post-K development, and the next gen HPC research. 4月1日より理研計算科学センター(R-CCS 旧AICS)に異動し、京やポスト京、更に次世代HPCをリードします。よろしくお願いします。
My dear friend ‘Dr. HPC’ Prof. Jack Dongarra wins the ACM Turing Award, equivalent to Nobel Prize in computer science! I was one of the endorsers and am sooooooooo happy for him as well as our field! Million congrats Jack!
Very happy & honored to be awarded the Medal of Honor with purple ribbon, award from the Emperor, for my long time contributions to advancing HPC . I thank deeply all my collaborators, students, supporters that this award represents
The ribbon finally arrives! It came in a nice lacquered box along with a certificate from the Emperor and an instruction manual to how to wear it at a ceremony. Thanks again to everyone for all your collaborations & support through the years. 紫綬褒章のメダルと賞状が届きました。
To quote myself: “future improvements are likely to focus less on CPUs and GPUs and more on enhancing the performance of the memory and the interconnect fabric. The most promising avenues for reducing energy consumption per data movement mainly lie in memory technologies and
The symbiotic relationship between
#AI
and
#HPC
will only deepen in the coming years:
@ProfMatsuoka
AI technologies will facilitate more rapid advancements in HPC, while supercomputers, in turn, will enable increasingly advanced AI algorithms
Because software, and its stable and growing ecosystem is the most important factor in general purpose (super)computers, and we cannot get x86 license, but we can with Arm which sells 10x more chips then x86.
日本ソフトウェア科学界会からフェローの称号を頂きました。計算機科学者として大変名誉であり嬉しい限りです。I was awarded the Fellow status from the Japan Society for Software Science and Technology. It is so fortunate to receive such a honorable status as a Computer Scientist.
In Denver now for
#SC23
. My first SC attendance was back in ‘93 and have attended every one of them since so this is my 30th anniversary. As I noted on my award talk last year I was wowed by Cray3 back then, and I still get wowed now!
Some people have quickly associated the new name with the work of Hokusai “富岳百景 Fugaku Hyakkeni” meaning “the hundred scenes of Mt . Fuji” but phonetically can be read as “a hundred K” meaning that Fugaku is 100 times faster than K on some target apps
PCs: Apple (M1), Microsoft (Surface); Servers&Clouds: Amazon (Graviton), Microsoft (Azure), Google, others (Ampere etc.); HPC: Fujitsu; Smartphones & Tablets: everyone. Now you know why NVIDIA intends to acquire Arm. For Intel, AMD is now a key ally to counter this massive flow